Projects

• Led a full rewrite of the client-facing frontend, improving performance, usability and platform adoption.
• Developed regional-scale data exploration and export tools that process years of data quickly.
• Moved static and spatial content delivery onto a CDN, improving caching efficiency and global latency.
• Architected and delivered online service ordering, integrating with existing operational and scheduling workflows.

Capture — led the refactor, redesign and production deployment of a business-critical field app, reducing support burden while improving release confidence, user experience and maintainability.

• Hand-coded advanced digitising features: performant survey gridding, manual digitising, and GPS-track-based digitising.

ATVise — led development on a cross-team project compiling location-aware in-field text, voice and image recordings into context-enriched, editable reports for agronomists and agents to share.

• Led the design and implementation of a centralised crop data platform, aligning and cleaning data across multiple legacy systems so that analytics and reporting could be trusted.
• Built an analytics API with DuckDB and tRPC, aggregating large and disparate Parquet datasets while still adhering to MyFarmWeb’s data-access authorization model.
• Redesigned the storage and processing of large spatial datasets, substantially reducing storage costs.
• Designed and built a bespoke PDF map-book templating engine, dramatically cutting turnaround from sample collection to delivery.

• Designed the internal packages, tooling and contribution constraints so that safety is enforced by the guard rails themselves rather than left to code review — opening the platform to LLM-assisted contributors.
• Built CI/CD pipelines and automated end-to-end testing across multiple frontend applications.
• Currently leading the migration off Firebase, which serves as the syncing layer for the Capture app.

Each thematic map is a mini-system in its own right, with its own data pipeline, caching layer and scheduled jobs behind it.

• Built entirely from publicly available datasets.
• Nothing is proxied live except select API tiles; the rest is ingested on a schedule rather than fetched on demand.
• Cached aggressively and selectively at every layer.

Co-founded a consultancy delivering production software for a range of clients, contributing to architecture, implementation and delivery while owning defined technical scope alongside engineers, designers and client stakeholders.